SPEt‘TRALSENSITIVITYOF THE FOVEAL CONE PHOTOPIGMENTS BETU'EEN 400AND 500nm' VNIAXXE C.
SUITH
and
JOEL
POKORN~
Eqe Research Laboratories. Universit) of Chicago. 950 E. 59th Street, Chicago. lllinoiz 60637, U.S.A.
Abstract-A comparison was made between the shape of the iodopsin absorption spectrum calculated for appropriate optical density to (I) a set of KGnig-type fundamentals in which the tritanopic copunctal point was set on the alrchne and (2) data obtained from red-green dichromats using high intensit! heterachromatic flicker procedures which eliminated participation bq the short-wavelength sensitive mechanism. The transformation of normal color mixture data resulted in two fundamentals which gave a reasonable prediction of the tritanopic coefficients. The dichromatic HFP data corrected individually to average macular pigment agreed with their respective fundamental above 130 nm. The HFP data and transformation wereconverted to a retinal level, quantized and plotted as a function of wavenumber. For the middle-wavelength-sensitive mechanism. the protanopic HFP data and its Kijnig-type fundamental agreed with the predicted absorption spectrum above 460 nm. The deviations below 160 nm had the shape of the lens absorbance curve. For the long-wavelength sensitive mechanism. the deuteranopic data and its KGnig-type fundamental agreed with the predicted absorption spectrum above 520 nm. The deviations belou 520 nnl could not be fit solely by, the lens absorbance factor used abobe. but needed in addition. added macular pigment of optical density at 460 nm of ca. 0.12. This result was checked by calculating predicted tritanopic coefficients for the two predicted absorption spectra, when the long-wavelength sensitive spectrum was screened by a slight amount (o.d. of 0.12 at 460 nm) of macular pigment. These predicted coetiicients agreed with the Wright tritanopic coefficients. We conclude (a) that the shape of the iodopsin absorption spectrum provides a reasonable basis for computation of absorption spectra of the middleand long-wavelength sensitive cone pigments and (b) that long-wavelength sensitive cones of deuteranopes. tritanopes, and normal trichromats are subject to a selective screening filter of optical densit! at 460 nm of 0.12 and spectral shape similar to macular pigment.
INTRODIJCTION In an earlier
paper
(Smith
and
Pokorny.
1972) we
showed that the spectral sensitivities of protanopes and deuteranopes, which mav be expressed as transformations of the CIE color mixture data, are similar in shape on their long-wavelength slopd to measured isolated visual pigment absorption spectra. It is our purpose to extend this analysis to a consideration of the spectral sensitivity of protanopes and deuteranopes in the wavelength range 4CO-530 nm. The wavelength range 400-520 nm poses a number of problems. (a) There is a problem in using protanopic and deuteranopic luminosities. The region 400-520 nm is a region where two photopigments are active. In this study we report protanopic and deuteranopic luminosity using heterochromatic flicker photometry on blue backgrounds at flicker rates above the fusion of the short-wave chromatic mechanism. (b) There are experimental and theoretical problems with the color mixture data. The experimental problem ’ This work was supported_by USPH National Institutes of Health. NE1 grant EYOO513.The major substance of this paper was presented at the CVS-OSA symposium in Rochester, October 1973. Y.R. I%?--*
concerns the exact nature of the color mixture data at short-wavelengths. The 1931 CIE color mixture data (Wyszecki and Stiles, 1967), the Stiles (1955, 1959) 2’ data, and Sperling’s (1938) data. when transformed into an equivalent primary system. differ in the amount of negative G primarT needed at short-wavelengths. Stiles (1955) and Sperlmg (1958) have commented on the variability,of matches at short-wavelengths. The theorettcal problem arises from the fact the 1931 CIE XYZ Standard Observer was set up with underestimated luminance below 460nm. Hence both the color matching functions and the spectral coefficients are incorrect at short-wavelengths. Judd’s (1951) modification involved adjusting the spectra1 coefficients o\er a wide spectral range in order to maintain standard source B at the same position in the W.D.W. coordinate system. This procedure does not guarantee the accuracy of the coefficients. To counterbalance these problems with the color mixture data. we have available an additional tool in the tritanopic color matching characteristics. Since small field fovea1 color mixture functions of color normal observers (Willmer and Wright, 1915) closely resemble those of tritanopes (Wright, 1952). due to the inert ocular plymrnts. In the lower panel are shown the predicted tritanopic color matching characteristics c:&zul;~ted u.ersers. The Vos and Walralen 119711primaries do not fit the Wright (1952) data. The Je~intions are greater than an! of Wright’s individual observers. These data suggest a problem u ith the Vos and Ovalraven I 1971Jprimaries. \Ve can con4der two possibilities:(a]The Vos and Walraven (1971) model is in error: (b) Judd’s (I 95 1) modi~cation of the CIE is in error. in the luminosity- curve and or in the coefficients. We turned first to the Vos and Walrnven (1971) model: their R and G functions are accurate above 520nm. Their choice of copunctal point for protanopes and deuteranopes gives a good empirical fit on the Iongwavelength slopes of the dichromatic functions. The tritanopic copunctal point is empirically derived from the tritanopic coeticients. and would be in error only if the revised SYZ diagram is in error. rl. possible source oferror in their model is their description of luminance. The! used a linear sum ofcontrihutions of the R. G and B functions. We should like to cowider the possibilit! that rhe s2lort-wavelenpthsensitive pigment makes no contribution to the CIE v;. function. This is not a new idea. Guth, Alexander. Chumbly, Gillman and Patterson (1968), for example have a model of color vision in which the CIE C>.function is called an achromatic Iuminanct: function and rtceikes no input from the short-uave sensiti\-e pigment. Guth and Lodge’s (197:) point that HFP should be regarded as .-achromatic” as compared ivith. sa?. brightness matching is based on the fact that HFP 1s obtained above chromatic fusion. A similar argument ma! be applied to the short-wavelength mechanism. bvhich is knotvn to have a low fusion frequency. With kolutcd II mrchaniws. Brindle>. DuCroz and Rushton I 1966) found ;I value of IS Hz for tv;o observers for the /I, mech;mism. Green (1969) has found Ktlues of 20-X Hz for the high frequcrq end for two observers at a luminance estimated at 500 td for the f/, mechnnibm. Since even low luminance tlicker photometrb Ina> be performed at IL-16 Hz at short-uakelengths. it is entireI> pocsihle that HFP data are obtained above rhe fusion of the short-~v~l~elen~th mechnniscn.
C’opunctalpolnr5 Tritanopic
Protanopc
Luminance SR, t SG, = y,. Solution equations SR, = +0~155l-Lf,, ~0.54311!t. . SG. = -0.1j511.fi,, 045654.?,, /I
+
-0.032861. + 0~031562:;.
We therelore calculated ;I yet of receptor characteristics. based on ;I jet of three copunctai points plus the proviso that the Y function represents the sum of the R and G functions. In this modei. the protanopic and Jeuteranopic copunctal points were the same as used by L’os and Walraven ( I971 ). The tritanopic copunctal point wx set on the alqchne. at .s, = 0.1745. Table I shows the equations. Figure 2 shows the tritanopic coefficients for the tr~~nsforin~~tion.The! are some%vhat improved over the coeficients obtained from the Vos and Walraven ( 1971) model. Hoivever. coefficients for wavelengths 170. 490-510 nm still lie outside the range of Wright’s (1952)obser\er+. Ifive move the tritan copunctal point to .x, = 0.1739. the fit improves somewhat. Of course with such ;I procedure. we are tnoving further From Thomson and Wright’s (1953) empiricnIl> determined copunctal point. We think that these remaining discrepancies result from errors in Judd’s (I951 I revised X12 di:tgram at short-vvacelengths. THE
Lli~IISOSITl’
D \T.\
Since the attempt to derive visual pigment coefficients from the normal color mixture data was not
Wavelenqth.
“tn
Fig. 3. Predicted tritanopic coefficients for the transformation given in Table I. The meaning of the symbols. solid lines. and vertical bars are the same as in Fig. I(b).
Spectral
sensitivity
completely successful. our next step was to determine luminosities on dichromats using high intensity-high flicker rate HfP uith blue backgrounds. a method pre\iousl> used by Miller (1972) and Smith and Pokorny (1973). to reduce the sensitivity of the short-waLelength mechanism. It has become clear to us that the contributions of the various cone outputs to the achromatic system in HFP ma) barb as a function of luminance. flicker rate. and the standard wavelength. Some of our pilot data suggest that normal observers and protanopes show no difference in HFP obtained using high luminance-high frequency procedures which eliminate possible short-wavelength receptor response and low luminance-low frequency procedures which allow the possibilit> of short-wavelength receptor rcsponse. The deuteranopic data are equivocal. For the present study we wished to ensure that onlb one \ isual pigment could contribute to the dichromats’ HfP thresholds.
The equipment is that described brietlh in a previous report (Smith and Pokorny. 1973). A two-channel system placed a Z mm image of a 75W xenon source at the plane of the observer’s pupil. The light from the source was collimated by a lens and then followed by a beam splitter. i\ variable speed single sector disc placed in the collimated beams allowed alternation of rhe two channels. In one channel (the variable tield) rhe disc was followed by a Iens’which focused the image of the arc in the plane of a two log unit metallic neutral densit: wedge. Light was recollimated. passed through a field stop limiting the field of biew to 2. and then focused on 2 mm artificial pupil. Light from the second channel (the standard field) remained collimated after passing the plane of the flicker disk until it was recombined *ith that of the first channel just before a field stop. Because one of the channels was inverted following the flicker disk. the replacement ofone channel b) the other was uniform. scanning from top to bottom. The neutral density wedge was driven b) a gear motor and the output of a directly coupled potentiometer was read on a voltmeter. The wedge could be controlled b) observer or experimenter. Fixed metallic neutral density filters Lvere also used. Interference filters were used to control spectral composition. A series of 13 filters were used with nominal pe;tks at -100.405. 410.41O.G6. 445. 453,464. JSO. 502. 510. 5SO and 65-l nm. At 8 x IO rectangular background field was probided by a third channel. using a tungsten iodide source and appropriate filters. A broad band filter (Kodak 47B1 was used in the background to depress the short-waLelength sensitive mechanism.
Calibrariom The interference and ND. fixed filters were calibrated using a Carey-14 recording densitometer. In addition, the neutral filters and the wedge were calibrated in position for each of the interference filters. For this task a United Detector Technology PIN IO photodiode was used. Three shortwavelength filters 400. 410, and 420 were three-cavit) filters (Ditric Optics) with half-band passes of 7-9 nm and visible side bands blocked to 10m6. The remaining interference filters (Schott) had half-band-widths of 10-15 nm.
165
of the fovea1 cone
The relative brated
ipectrul
b! Optronics
response: of the PIN Laboratories.
used to check the relative
radiance
filters on each experimental day. An estimate of retmnl illuminance
diode
The calibrated through
was calidiode ua>
the interference
was made placing
an
llford SEI exposure photometer in the plane of the artificial pupil ior the standard tield with the 578 interference filter. The obr.tinsd measurement of retinal illuminance was YOO td. The retinal illuminance of rhs background \\as ascertained first bb establishing a HFP match for the 178 filter to the snndard illuminance and then b> direct homochromatic brig.htncss matching of the bariable field to the upper portion ot the background. The retinal illuminance of the background uas 500 rd. Sllhjrcrs The Jeuteranopes and protanopes were obtained as a result of advertising in the Unitersit) of Chicago student newspaper. All sere males. The) ranged in age from 17-31 with an average age of35. .AII observers were screened using A0 HRR plates. Ishihars pwudoisochromaric plates. the Farnsworth Munxll loo-hue test. and the Nagel anomaloscope. OnIF those observers who could match in color and brightness both rhe red (670 nm) and the green (536 nm) to the )ello\\ (5S9nm) on the Sage1 anomaloscope u-ere accepted as dichromuts. All observers had normal corrected acuit> and used rheir oustomarb refractive correction during the stud!.
At ths {tart of each session ths observer adjusted the chin rest until his eye was centered on the artificial pupil. The obserxsr then dark adapted for 5 min. Hetcrochromatic flicker photometry was performed to a 540 nm standard for protanopes and to a 580 nm standard for deutsranopes. The test walelengths were presented in rnndom order once in each session. The observer adjusted the radiance of the inriable test wavelength until there was a minimal or absent sensation of Hicker. The skperimenter then adjusted the radiance of the test wavelength in small steps around the observer’s setting in order 10 establish a match range. The Hicker rate (X-27 Hz) was adjusted to maintain the match range at 0-O-I log unit of radiance. After an initial practice session. the procedure v.as repeated on three separate sessions. The data reported are the average data for the three sessions.
Rewir.~
Figure 3 shout the log relative energ! as a function ofwa\elcngth for four deuteranopes (louer panel). and five protanopes (upper panel,. The solid lines show the relative sensititities of the transformations given in Table 1. The deuteranopic points scatter widei) across the predicted line. The protanopic points tend to lie above the prediction. Inset in each panel is the densit! difference betueen an HFP match at 54Onm and 464nm at 0 and 9 . i.e.: an estimate of macular pigment densit! for each observer. The protanopes shorted less \ariabilit\. in their macular pigment, ;I tinding apparent also in Pitt’s (1935) data. The estimated macular pigment density at 464 run ranged from 0.136 to 0% in the deuteranopes and from 0.263 to 0375 in the protanopes. The estimation may well be an underestimate since according to Polqak ( 1941) the
points tend to cluster above the predictions made by the transformations. the transformations anti part Bv definition. ium&sities of the normai iuminositj; iunction. These part tuminmities are, of course, determined by Judd’s (1951) revised luminosity curve. The Judd (1951) revised luminosity curve fits normal NFP data well above 430nm. The corrections made by Judd were based on the data of Wald (19453 (absolute threshold), Weaver (1949), Thompson (1949). and Ishak (192) (brightness ~at~h~n~ at short-wavel~ng~hs~ and Gibson and Tyndah (19131 ~extrapolatjon~. We have noticed small but consistent deviakms of the HFP data of Coblenz and Emerson (19 171, Sperling ( 19581, and Wagner and ~oynton (1971) from Judd’s (1951) revised fuminosity function at wavelengths below 430 nm. The Judd { 1% If revised luminosity function underestimates HFP luminosity (average of studies cited above) at wavetengths 4OO-430 nm. the difference being 0.31 log unit at 400 nm. 0.29 at 410 nm, 0.20 at 420 nm and @15 at 430 nm. Figure j shows the average of the dichromatic data plotted as log relative quanta1 sensitivity as a function
Fig. 3. Log reiative energy as a function of wavelength for five protanopes performing hererochromatic flicker photometry to a 540 nm standard [panel (a)] and 4 dcuteranopes [panel (b)] performing heterochromatic flicker photometry to 8 580 nm standard. The solid lines are the relative sensitivities of the transformations given in Table i. Inset in each panel is a ~omparis& of the match of 540 nm to 464 nm made with foveai iiaation and in the 9” vertical meridian. Symbols for panei [a): Protanopie observer RS O; MS V; DP A: K.a 22;MG 0. Symbols for panel(b): ~uterano~ic observer M3 0: KC Cl: NW V; fC 0.
macular pigment may extend to 17’. Further, the procedure implies unpravable (and improbable) assumptions of the identity of optical density, distribution, and sensitivity of the cones at the fovea and 9’ vertical mer-
idian. The same problems however are evident in all psychophysical techniques of estimating macular pigment. The measurement does pravide us with a method ofreducing observer variability due to interindividual variation in macular pigment density. Figure 4 shows the data corrected for each observer to a macular pigment of cl,,, 0.53. using the Wyszecki and Stiles f 1967) spectral correction. The solid lines are again the transformations shown in Fig 3. They provide a reasonable description of the data except b&w 430nm. For wavehzngths below 430 nm the data
167
Spectral sensitivity of the fovea1 cone
Wovenumber.
cm“
-2.21 25.300
20,000 Wavcnumbef.
L5.000
cm-”
Fig. 5. Log relative quantai sensitivity as a function of wavenumber for protanopes [panel (a)] and deuteranopes [panel (b)]. The dashed lines show the relative sensitivities ofthe transfor~tionsofTable I corrected to a retinal level. The solid lines are the predicted visual pigment absorption coefficient as given in Fig. 1. The inset in each panei shows the difference of the psychophysical estimates from the predicted visual pigment absorption coeffcient. The solid line in the insets is a factor for added lens density, the dashed line on the inset of panel (b) is a factor for added macular pigment. Symbols: average data of Pitt A; Hsia and Graham El; and the present study 0; all corrected to a retinal level. The crosses on the inset of panel (b) represent the difference between the transformation and its predicted visual pigment absorption coefficient.
of wavenumber. The data were corrected for an average lens and macular pigmentation ((i,,, of 053) as tabulated in Wyszecki and Stiies (1967). We have added two extra sets of dichromatic data: (a) Pitt’s ( 1935) protanopic luminosity data were added since in ‘The Pokorny and Smith study allows a comparison of the three tasks (absolute thresholds. HFP to 5SOnm and HFP to 3OOO’K)for five observers (two protanopes and three deutzranopes). The flicker rate below 500 nm was 10I2 Hz at 40 td. There was no difference in the thresholds as a function of method. The data are not replotted for reasons including: (a) we have no independent assessment of macular pigment and too few observers to justify averaging; (b) we could not measure sensitivities at wavelengths betow 420 nm: (c) examination of the long-wavelength slope suggests that our calibration procedure was not as precise as that used for this study.
our laboratory we can detect no difference in high and low frequency HFP functions for protanopes. ib) The Hsia and Graham (1957) absolute threshold data were added since Pokorny and Smith (1972) reported that using the Hsia and Graham (1957) procedure, absolute thresholds yielded identical results to HFP for dichromats.’ although other procedures (Guth and Lodge, 1973) do not show this identity. The Ikeda (1963, 1964) data suggest that stimulus duration is the determining parameter. The data for the Hsia and Graham (1952. 1957) and Graham and Hsia (1969) procedure (4 msec stimulus flash) suggest that their absolute thresholds follow the absorption coefficient of the most sensitive of the visual pigments. For protanopes and deuteranopes their respective long-wavelength-sensitive pigments appear to be the most sensitive over the full wavelength range, and may thus be compared to the same pigments isolated in the achromatic system by high-flicker rate HFP. The dashed lines show the relative sensitivity of the transformations of Table I corrected to a retinal level. The soIid lines are the proposed visual pigment absorption spectra based on the shape of the iodopsin absorption spectrum (Wald. Brown and Smith, 1955) calculated for appropriate optical densities. The spectra are slid on the horizontal axis to provide a good fit to the data and transformations on the long-wavelength slope with the result that the spectrum compared with the protanopic functions has a L,,, CCI. 532 nm and that compared with the deuteranopic functions has a i.,,, cu. S55 nm. These estimates for i,, are relative to the accuracy in i,,, of the iodopsin spectrum. The data points and the transformations fit the proposed visual pigment spectrum for wavelengths above 520 for deuteranopes and above 460 nm for the protanopes. There is a notch at 490 nm in the data, evident for both classes of observer, which may be an overcorrection of macular pigment at that wavelength (cf. vos, 1972). The deviations of the dichromatic data and functions from the proposed visual pigment spectrum become severe at short-wavelengths. inset in the graphs are the density differences of the dichromatic data points from the proposed visual pig.ment spectrum. The solid line through these points IS a correction of the lens density factor to 1.333 times the Wyszecki and Stiles (1967) tabulations. The solid line provides a good fit of the protanopic deviations. but not of the deuteranopic deviations. The dashed line on the deuteranopic inset which does provide a reasonable fit to the deuteranopic deviations is obtained by adding increased macular pigment of 0.12 optical density at 460 nm to the lens correction. DISCUSStOS
The results show that the dichromatic data and the transformation of the CIE data in the wavelength range 43~7~nm approximate the visual pigment absorption coefficient predicted from the iodopsin
Jle-waxiength cide a good iii TV> the Wri@t (1351) tritanopic coefficients. The c~~etti&nts at 520 nm were checked at :I number af values +I :tddrd macular pi~msnt at 460 nm. The preof,1 dictions fX within U.right’\ t 19521obxeriers proI iJed the 4en4t! \rf mitcuktr pigment Aied to the tongwa~elen&t ~2nsirii~ pighrtnr ha5 optical densit! at I60 nm betuecn 0419 timi 0.16. We thus conclude that the io&)psin spectrum when ndjuqted to appropriate densit) and slid MI the wa\enumber axis can pro\ ide a re;tsonable description of the L~YAJ:II pipmrnt absorption spectra of the mAidleand long-ua\elen$h sensitiie fove:tl cones although SC cspcct min~x deviations .tt the short-walelengths. Further. we come to the conclusion that we must ahon ditl’erent r\rntxints of macul;tr pigment for the two
Fig. 6. Panel (a) Log relative quanta1 sensitivity of the proposed human visual photopigments. The symbafs and lines correspond to the continuous lines of Fig, 5. Symbols: Middle-wavelength sensitive 0; long-wavelength sensitive a. Panel (b) Tritanopic coefficients predicted by the pigment absorption spectra of panel (a). The symbols are the same as those in Fig. I(b).
absorption spectrum (,at appropriate opticat densit!) prl” idsd the lens correttion factor is increased 01; RI. I.333 and that ;I diirerrtntial macular pigment correction is applied to the middle- and lon~wavelen~th sensitive mechanisms. There is no indic~ltion that tluorescence of the optic media or retina has contribrtted to
the measured sensitivities. The data do not show any heightened sensitivity that might be caused b! Huorescence. The added optic media correction factor is that predicted b\ C’oren and Girgus (1971) for JO-.vr-olds and agrees with Tan’s (1971) data for the ape group X-19. Although such a correction is necessarily CR/hoc, it is a reasonable correction. given our knowledge of the absorption in the ocular media. The macular pigment correction presents a different problem: it appears that the deuteranopes have slighti! sreater average macular pigment. The same result oc’curs in the transform~~tion of the revised CIE color mixture data, a result which preserves the inteprirr of the copunctal points. The long-wavetength sen4&e cones of normal trichromats are apparently subject to a 4ightly higher macular pigment than the mid-
Fig. 7. The effect of adding a selective Biter to the proposed Lang-~vave~~n~th sensitive pigment. Panel (a) Pigment absorption curves of Fig. 6 with a macular pigment filter of o.d. @il at 460 nm screening the long-wavelength sensitive photopigment. Panel (b) Ttitanopic coef%5enrs predicted by the spectral sensitivities shown in PaneI (a). The symbols are the same as those in Fig. i(b).
Spectral sensitivity of the iovcal cone
classes of cones: namely, maximal optical density of about 0.5 for middle-wavelength sensitive and @62 for long-wavelength icnsitice cones. This suggestion is not contradicted by the available psychoph>&tl measurements on normal macular pigment. since the predicted appxent average macular pigment density in normals given d,,,, of 0.5 for middle- and 0.61 for long-wavelength \sn\itivs cones would be about 057 at 460 nm. However. since the macular pigment is described as lying around the fibers of Henle (Segal, 1950).primarily in the outer plexiform layers (Gass, 1973). or through layers 4-9 of the retina (Polyak. 1941) the assumption of differential pigment density affecting the two classes of cones i4 somewhat puzzling. A number of alternatives may be considered. (a) A larger differential optical density of the two photopigments: In order to fit the deutan data to a hypothetical pigment absorption curve, we would need to compute it for dilute optical density. This possibility is contradicted by the results of Miller (1972) and ourselves (Smith and Pokorny. 1973) who find the longwavelength sensitive pigment to be in density 04-0.55. Further. our calculations of predicted tritanopic coefficients based on a larger optical density differential between the two pigments do not fit the Wright (1952) tritanopic data below 440 nm. (b) Two pigments in longwavelength sensitive cones: Two pigments each absorbing quanta should look like a pigment with apparently greater sensitivity than the proposed absorption spectrum. We are concerned with a sensitivity which lies below the proposed absorption spectrum. (c) Intrusion of the short-wavelength mechanism by inhibitory interaction with the long-wavelength mechanism. Such a hypothesis requires that some proportion of normal short-wavelength sensitive receptors exist whose output consists of an inhibitory linkage to the long wavelength sensitive receptors. These shortwavelength sensitive receptors would also be postulated for tritanopia-a possibility since there is no conclusive evidence that tritanopia is a pigment defect. However, our calculations show that predicted tritanopit coefficients for such a system do not agree with Wrights (1952) tritanopic data. (d) A stable photoproduct of the long-wavelength sensitive pigment acting as a screening pigment: This notion was suggested by Dartnall(1962) for the case of indicator yellow as a screening pigment for rhodopsin. The concept was carefully examined by Goldstein and Williams (1966) who concluded that the low optical densities in human rods and cones preclude stable photoproducts from acting as effective screening pigments. They worked however with Rushton’s (1956) value of 0.15 optical density for rods. We allow higher optical densities of 0.3-04. Even so, a substantial bleach (probably- 50 per cent) would be required to build up an appropriate density of any hypothetical screening photoproduct for long-wavelength sensitive cones. even given optimal conditions for such build-up. Our analysis indicates that the selective filtering of the
169
long-wavelength sensitive cones occurs at luminances ranging from absolute fovea1 threshold to SO0td. (e) .-\ substance of spectral shape similar to the macuhr pigment which is present in the structure of the cones and mav occur in both peripheral and fovea1 cones: The relatively higher dcnsitv of such a substance in the long-wavelength sensitive cones may reflect a greater length of the outer segment, a result which we might expect based on the higher density of the long-wavelength sensitive cones found by Miller (1972) and Smith and Pokornl (1973). In their estimation of the optical density ol’ human cones. Dobelle. Marks and YlacXchol (1969) imply that the optical densit) ivill be proportional to the length of the outer segment. (f) There is the possibility of error in the Wald. Brown and Smith ( 1955) measurements. At 500 nm the percentage extinction for iodopsin is 50 per cent and we would need to invoke a IO per cent error in Wald. Brown and Smiths (1955) spectrum. The two estimates using independent techniques made by Wald. Brown and Smith ( 1955) of iodopsin absorption at 500 nm differed by only 2 per cent. (g) A remaining alternative would be to state that the deuteranopic curve is the true absorption curve and that the protanopic curve results from a summation such as mentioned in (b) above. Such a hypothesis predicts incorrect tritanopic coefficients. Further we would need to accept that the similarity in shape of the protanopic functions to the iodopsin spectrum is coincidence. We prefer to think that the shape similarities found between the visual pigments of many species in the animal kingdom does extend to the human cone photopigments. Of these alternatives, the most likely seems to be (d) and (e). which of course are essentially ad ltoc explanations. Thus vve come to the conclusion that our analysis requires a selective screening of long-wavelength sensitive comes by a filter whose spectral transmisston is similar or identical to macular pigment. Guild (1931). Wright (1923) and Stiles (1955) found considerable variab&ty- in the color matches of normal trichromats for the spectral region 470-500 nm. W’right (19X3) points out the different color matches esceed observdtional error and may indeed reflect receptor differences. From our point of view such variability might reflect variation in differential density of the screening pigment. Further. it appears that the human cone visual pigments are closely similar in shape to the Wald Brown and Smith (1955) measured absorption spectrum for iodopsin. However we would not be surprised by minor deviations in the spectral region 400450nm. the region of greatest uncertainty in both Wald Brown and Smiths (1955) data and the psychophysics. We conclude that the shape of the iodopsin absorption spectrum provides a basis for calculation of absorption spectra for the middle- and long-wavelength sensitive human cone photopigments.
REFEREhCES
Xlpcrn 41.. Xlaaseidvaag F. and Ohba N. ( IYTI) The kine tics of cone visual pigments in man. t’isioft Rrs. 11. j39549. Boettner E. A. and VVoltet 5. R. I 1962)Transmission of the ocular media. Inrrsr. Ophthai. 1, 776-783. Bone R. A. and Spsrrock J. 51. B. (1971J Comparison of macular pigment densities in human eyes. C’isio/tR~s. 11. IOj7- 106-t. Brindle! G. S.. Du Croz J. .I. 3rd Rushton W’.A. H. (19661 The tlicker fusion frequent)- of the blue-sensitive mechanism of colour vision. j. Ph_Mffi, Lond. 183, 497500. Brolin S. E. and Cederlund C. (1953) The fluorescence of the lens of the eves of different species. Actrr ophrhal. 36, 37c32s _
.
Coblentz W. W. and Emerson W. B. (1915) Relative sensibilitv ofthc average eve to lizht of different coIors and some p&tical applications to radiation problems. Bttil. n~trz. Bur. Sron&rinms1-l. 167-236. Collins F. D., Love R. M. and \lorton R. A. i 1952) Studies in rhodopsin-IV: Preparation of rhodopsin. Biochrm. J. ;I, 192-29s. Coren S. and Girgus J. S. (1972). Density of human lens pigmentation: irr rirr) measures over an extended age ranye. t-isiuct Rcs. 12, %.r-316. Dartnail H. J. A. 113621 The photobiology of Visuai Processes. In Tile Ertrdited by Davson H.), Vol. II. pp. ??I533. Academic Press, New York. Dobellc W. H.. Xlxks W. B. and MacNichol E. F. (19691 Visual pigment density in single primate fovea1 cones. Scit%ce. .\‘. y. 166. IbOS-I 510. Gass 1. D. M. (19-2) Nicotinic acid macttlopathy. .-lrrr.J. ~~~ir~~~~. 76. SO&5 Ii). Gibson K. S. and Tyndall E. P. T. i 1923) Visibility of radiant energy. Scienr. PHI?.Birr. Srund. 475, 131-191. Goldstein E. B. and WXiams T. P. (1966) Calculated effects of “screening pipm:nts”. Vision Res. 6, 39-50. Graham C. I-f. and Hsia Y. (1969) Saturation and the fovcal achromatic interval. d. opr. SM. .Arn.59,993-997. Green D. G. (1969) Sinusoidal fiicker characteristics of the color-sensitive mechanisms of the eye, Vision Rrs. 9, 591601. Griitzner P. and Kohlrausch A. (1961) Der lichtwrlust in der macula lutsa und das farbensehen von deuteranomalen. PfIiigerr Arch. ges. Ph,vsiof. 274, 3 I S-330. Guild J. (1931) The calorimetric properties of the spectrum. Phi!. Trtrrts. R. SOC. 4 230, 149 157. Guth S. L.. Aieuander J. V.. Chumbly J. I.. Gillman C. B. and Patterson hf. 41. (196s) Factors affecting luminance additivity at threshold among normal and color-blind subjects and elaborations of a trichromatic-opponent colors theory. I-isinn Res. 8, 913-925. Guth S. L. and Lodge H. R. (19733Heterochromatic additivity. fovenl spectral sensitivity and a new color model. f. opt. Sac. .-bn. 63. -tit%462. Hsia Y. .rnd Graham C. Ii. (1953)Spectral sensitivity of the cones in the dark adapted human eye. Proc. mm. Acud. Sci. 38, SC-Sj. Hsia Y. and Graham C. H. (1957) Spectral luminosity cur\es for nrotanooic. deuteranopic and normal subjects. Proc. n40;. ilcad. Sci. 13. 101I-1019. Ikeda M. (1963) Study of interrelations between mechanisms at threshold. J. opt. Sue. Am. 53, 130j_f313. fkcda M. (1964) Further use of the summation index for the study of color vision. J. opt. Sot. ,Arrz.S4, 89-9-k
Ishak I. G. H. I ii)jZ) The photopic lummosity curve for a group oftiftsrn Egyptian trtchromats. J. clpr. SOC. im. 42. 519-j;!,
Judd D. 3. I 195:JSeer. Rep. Coiorimetry and artificial daylight. In Proc. 13~11Srssip~t CIE: Stockholm Vol. I. Tschn. Comm.. No. -. Klanp G. (1%) Measurements and studies of the Huorsscrnce of the human lens in ciro. .&m ophthctl. Suppl. 31. I-Ii. Lr Grand Y. I IY3S) Sur la tiuorsscence du cristailin. C.r+. .k‘Zif. Sci. zu:.
11zs- I 130.
L. Grand Y. (19fi) Nou~clles rzcherchrs sur la fluorescence du cristallin. C.r. ,fcnd. Sci. 225. 1031-10~2. Ludvigh E. and McCarthy E. F. (1939) Absorption of visible light by refractive media of the human eye. .-l~hs. Opitthai. 20. zi-51.
Miller S. S. (1973)Psychophysical estimates of visual pig. menr densities in red-gcen dichromats. J. Plt.~siof.. Lmf. 223. Y9- 107. Pitt F. H. G. (1935) Characteristics of dichromatic vision, with an appendix on anomalous trichromatic vision. Spec. Rep. wd.
Rex Council 200, I-3.
Pokorny J. and Smith V. C. (1972) Luminosity and CFF in oenteranopss and protanopes. J. opt. Sot. .irn. 62, 1I I-1 17. Polyak S. L. (IY-IL)The Rrritra. L’niversitv of Chicago. Ruddock K. H. 11963) Evidence for macttlx pi~menmtion from colour matchine data. lisi~n Rzs. 3. ili-429. Rushton Et’. ,A. H. (1956) The Rhodopsin density in the human rods. J. Ph,vsiof.. Lad. 134, 306. Said F. S. and Weals R. A. (1959) The variation with age 01 the spectral transmissivity of the living human crystalline lens. Grr-ortro/ogici3, 3 I i-33 I. Ssgal J. I 1950)Localisation du pigment maculairs de la rCrinr. Cx. &rrc. Sot. &of. I-M, Pt. 2, 163&1631. Smi:h V. C. and Pokorny J 11972) Spectral sensitivity of color-blind observers and the cone photopigments. t’isioh Rrs. 1’-9 203-2071.
Smith V. C. and Pokornl J. (1973) Psychophysical estimates of optical densit\ in human cones, C’isiortRrs. 13, I l99-
12oi. Sperling H. G. I 19%) An experimental investigation of the relationship between colour mixture and luminous et%c&x,-. In Prot. Yrh !VPL Prohte,ns 0s C&t-. (Symposium held at the Sarional Physical Laboratory. September 1957) pp. 251-‘79. Her Majesty’s Stationery Office, London. Stiles IV. S. t 1955) Interim report to the commicsion intsrnationale de l’eclarrage, Zurich. 1955, on the Xational Physical Laboratory’s investigation of ~olour-matching (1954). with an appendix by W. S. Stiles and J. 41. Burch. Dprica .Jcm 2, 16%ISI. Stiles VV.S. and Burch J. >I. (1939) N.P.L. colour-matchinp in\sstigation: final report (19SS). Opricu -Icr,t 6, l-26. Tan K. E. W. P. (1971) Vision in the ultraviolet. Doctoral thests. Utrecht. Thomson L. C. 119-19)Shape irregularities in the equal energy- Ium~nosity curve. Prof. Phys. Sot.. fmd. B62. 757?I?,
Thomson L. C. and Wright IV. 5 (1953) The convergence of the tritanopic confusion loci and the derivation of the fundamental response functions. J. opt. Sot. .-lm.13, 890s94. Trendelenberg W. E. T. (1961) Der Grsichrsinrt. Grund:ii~gr der ph~siologi~c~re}t opiik. 2nd Edition (edited by Monje M.. Schmidt I. and Schlitz E.). pp. 67-69. Sprinpr-Veriag, Berlin.
Spectral sensitivity of the fovea1 cone Vos J. J. (1973) Literature review of human macular absorption in the visible and its consequences for the cone receptor primaries. Inst. Percept. RVO-TN0 Report So. 1ZF (972-l 7. Soesterberg. Vos J. J. and Walraven P. L. (1971) On the derivation of the fovea1 receptor primaries. Vision Res. 11. 799-818. Wagner G. and Boynton R. M. (1972) Comparison of four methods of heterochromatic photometry. J. opt. Sot. .-lm. 62, IjOS-1515. Wald G. (1945) Human vision and the spectrum. Science, S.l’: 101. 653-658. Wald G.. Brown P. K. and Smith P. H. (1955) Iodopsin. J. gen. PhysioL 38, 623-681. Weale R. A. (1954) Light absorption by the lens of the human eye. Optical .-lcta 1. 107-l 10.
171
Weaver K. S. (L949)A provisional standard observer for low level photometry. J. opt. Sot. ,4/n. 39, 278-39 1. Willmer E. N. and Wright W. D. (1945) Colour sensitivity of the fovea centralis. .Vaturr. Land. 156, 119-L21. Wright W. D. (1929) A re-determination of the trichromatic coefficients of the spectral colours. Trans. opt. Sot. U.K. 30, 141-164. Wright W. D. (1951) The visual sensitivity of normal and aphakic observers in the ultra-violet. Ann. Ps@ol. 50, 169-177. Wright W. D. (1952) The characteristics of tritanopia. J. opt. Sot. Am. 12. 5G9-521. Wyszecki G. and Stiles W. S. (1967) Color Science: Concepts and Methods, Quantitative Data and F~rmwlas. John Wiley, New York.
SUITH
and
JOEL
POKORN~
Eqe Research Laboratories. Universit) of Chicago. 950 E. 59th Street, Chicago. lllinoiz 60637, U.S.A.
Abstract-A comparison was made between the shape of the iodopsin absorption spectrum calculated for appropriate optical density to (I) a set of KGnig-type fundamentals in which the tritanopic copunctal point was set on the alrchne and (2) data obtained from red-green dichromats using high intensit! heterachromatic flicker procedures which eliminated participation bq the short-wavelength sensitive mechanism. The transformation of normal color mixture data resulted in two fundamentals which gave a reasonable prediction of the tritanopic coefficients. The dichromatic HFP data corrected individually to average macular pigment agreed with their respective fundamental above 130 nm. The HFP data and transformation wereconverted to a retinal level, quantized and plotted as a function of wavenumber. For the middle-wavelength-sensitive mechanism. the protanopic HFP data and its Kijnig-type fundamental agreed with the predicted absorption spectrum above 460 nm. The deviations below 160 nm had the shape of the lens absorbance curve. For the long-wavelength sensitive mechanism. the deuteranopic data and its KGnig-type fundamental agreed with the predicted absorption spectrum above 520 nm. The deviations belou 520 nnl could not be fit solely by, the lens absorbance factor used abobe. but needed in addition. added macular pigment of optical density at 460 nm of ca. 0.12. This result was checked by calculating predicted tritanopic coefficients for the two predicted absorption spectra, when the long-wavelength sensitive spectrum was screened by a slight amount (o.d. of 0.12 at 460 nm) of macular pigment. These predicted coetiicients agreed with the Wright tritanopic coefficients. We conclude (a) that the shape of the iodopsin absorption spectrum provides a reasonable basis for computation of absorption spectra of the middleand long-wavelength sensitive cone pigments and (b) that long-wavelength sensitive cones of deuteranopes. tritanopes, and normal trichromats are subject to a selective screening filter of optical densit! at 460 nm of 0.12 and spectral shape similar to macular pigment.
INTRODIJCTION In an earlier
paper
(Smith
and
Pokorny.
1972) we
showed that the spectral sensitivities of protanopes and deuteranopes, which mav be expressed as transformations of the CIE color mixture data, are similar in shape on their long-wavelength slopd to measured isolated visual pigment absorption spectra. It is our purpose to extend this analysis to a consideration of the spectral sensitivity of protanopes and deuteranopes in the wavelength range 4CO-530 nm. The wavelength range 400-520 nm poses a number of problems. (a) There is a problem in using protanopic and deuteranopic luminosities. The region 400-520 nm is a region where two photopigments are active. In this study we report protanopic and deuteranopic luminosity using heterochromatic flicker photometry on blue backgrounds at flicker rates above the fusion of the short-wave chromatic mechanism. (b) There are experimental and theoretical problems with the color mixture data. The experimental problem ’ This work was supported_by USPH National Institutes of Health. NE1 grant EYOO513.The major substance of this paper was presented at the CVS-OSA symposium in Rochester, October 1973. Y.R. I%?--*
concerns the exact nature of the color mixture data at short-wavelengths. The 1931 CIE color mixture data (Wyszecki and Stiles, 1967), the Stiles (1955, 1959) 2’ data, and Sperling’s (1938) data. when transformed into an equivalent primary system. differ in the amount of negative G primarT needed at short-wavelengths. Stiles (1955) and Sperlmg (1958) have commented on the variability,of matches at short-wavelengths. The theorettcal problem arises from the fact the 1931 CIE XYZ Standard Observer was set up with underestimated luminance below 460nm. Hence both the color matching functions and the spectral coefficients are incorrect at short-wavelengths. Judd’s (1951) modification involved adjusting the spectra1 coefficients o\er a wide spectral range in order to maintain standard source B at the same position in the W.D.W. coordinate system. This procedure does not guarantee the accuracy of the coefficients. To counterbalance these problems with the color mixture data. we have available an additional tool in the tritanopic color matching characteristics. Since small field fovea1 color mixture functions of color normal observers (Willmer and Wright, 1915) closely resemble those of tritanopes (Wright, 1952). due to the inert ocular plymrnts. In the lower panel are shown the predicted tritanopic color matching characteristics c:&zul;~ted u.ersers. The Vos and Walralen 119711primaries do not fit the Wright (1952) data. The Je~intions are greater than an! of Wright’s individual observers. These data suggest a problem u ith the Vos and Ovalraven I 1971Jprimaries. \Ve can con4der two possibilities:(a]The Vos and Walraven (1971) model is in error: (b) Judd’s (I 95 1) modi~cation of the CIE is in error. in the luminosity- curve and or in the coefficients. We turned first to the Vos and Walrnven (1971) model: their R and G functions are accurate above 520nm. Their choice of copunctal point for protanopes and deuteranopes gives a good empirical fit on the Iongwavelength slopes of the dichromatic functions. The tritanopic copunctal point is empirically derived from the tritanopic coeticients. and would be in error only if the revised SYZ diagram is in error. rl. possible source oferror in their model is their description of luminance. The! used a linear sum ofcontrihutions of the R. G and B functions. We should like to cowider the possibilit! that rhe s2lort-wavelenpthsensitive pigment makes no contribution to the CIE v;. function. This is not a new idea. Guth, Alexander. Chumbly, Gillman and Patterson (1968), for example have a model of color vision in which the CIE C>.function is called an achromatic Iuminanct: function and rtceikes no input from the short-uave sensiti\-e pigment. Guth and Lodge’s (197:) point that HFP should be regarded as .-achromatic” as compared ivith. sa?. brightness matching is based on the fact that HFP 1s obtained above chromatic fusion. A similar argument ma! be applied to the short-wavelength mechanism. bvhich is knotvn to have a low fusion frequency. With kolutcd II mrchaniws. Brindle>. DuCroz and Rushton I 1966) found ;I value of IS Hz for tv;o observers for the /I, mech;mism. Green (1969) has found Ktlues of 20-X Hz for the high frequcrq end for two observers at a luminance estimated at 500 td for the f/, mechnnibm. Since even low luminance tlicker photometrb Ina> be performed at IL-16 Hz at short-uakelengths. it is entireI> pocsihle that HFP data are obtained above rhe fusion of the short-~v~l~elen~th mechnniscn.
C’opunctalpolnr5 Tritanopic
Protanopc
Luminance SR, t SG, = y,. Solution equations SR, = +0~155l-Lf,, ~0.54311!t. . SG. = -0.1j511.fi,, 045654.?,, /I
+
-0.032861. + 0~031562:;.
We therelore calculated ;I yet of receptor characteristics. based on ;I jet of three copunctai points plus the proviso that the Y function represents the sum of the R and G functions. In this modei. the protanopic and Jeuteranopic copunctal points were the same as used by L’os and Walraven ( I971 ). The tritanopic copunctal point wx set on the alqchne. at .s, = 0.1745. Table I shows the equations. Figure 2 shows the tritanopic coefficients for the tr~~nsforin~~tion.The! are some%vhat improved over the coeficients obtained from the Vos and Walraven ( 1971) model. Hoivever. coefficients for wavelengths 170. 490-510 nm still lie outside the range of Wright’s (1952)obser\er+. Ifive move the tritan copunctal point to .x, = 0.1739. the fit improves somewhat. Of course with such ;I procedure. we are tnoving further From Thomson and Wright’s (1953) empiricnIl> determined copunctal point. We think that these remaining discrepancies result from errors in Judd’s (I951 I revised X12 di:tgram at short-vvacelengths. THE
Lli~IISOSITl’
D \T.\
Since the attempt to derive visual pigment coefficients from the normal color mixture data was not
Wavelenqth.
“tn
Fig. 3. Predicted tritanopic coefficients for the transformation given in Table I. The meaning of the symbols. solid lines. and vertical bars are the same as in Fig. I(b).
Spectral
sensitivity
completely successful. our next step was to determine luminosities on dichromats using high intensity-high flicker rate HfP uith blue backgrounds. a method pre\iousl> used by Miller (1972) and Smith and Pokorny (1973). to reduce the sensitivity of the short-waLelength mechanism. It has become clear to us that the contributions of the various cone outputs to the achromatic system in HFP ma) barb as a function of luminance. flicker rate. and the standard wavelength. Some of our pilot data suggest that normal observers and protanopes show no difference in HFP obtained using high luminance-high frequency procedures which eliminate possible short-wavelength receptor response and low luminance-low frequency procedures which allow the possibilit> of short-wavelength receptor rcsponse. The deuteranopic data are equivocal. For the present study we wished to ensure that onlb one \ isual pigment could contribute to the dichromats’ HfP thresholds.
The equipment is that described brietlh in a previous report (Smith and Pokorny. 1973). A two-channel system placed a Z mm image of a 75W xenon source at the plane of the observer’s pupil. The light from the source was collimated by a lens and then followed by a beam splitter. i\ variable speed single sector disc placed in the collimated beams allowed alternation of rhe two channels. In one channel (the variable tield) rhe disc was followed by a Iens’which focused the image of the arc in the plane of a two log unit metallic neutral densit: wedge. Light was recollimated. passed through a field stop limiting the field of biew to 2. and then focused on 2 mm artificial pupil. Light from the second channel (the standard field) remained collimated after passing the plane of the flicker disk until it was recombined *ith that of the first channel just before a field stop. Because one of the channels was inverted following the flicker disk. the replacement ofone channel b) the other was uniform. scanning from top to bottom. The neutral density wedge was driven b) a gear motor and the output of a directly coupled potentiometer was read on a voltmeter. The wedge could be controlled b) observer or experimenter. Fixed metallic neutral density filters Lvere also used. Interference filters were used to control spectral composition. A series of 13 filters were used with nominal pe;tks at -100.405. 410.41O.G6. 445. 453,464. JSO. 502. 510. 5SO and 65-l nm. At 8 x IO rectangular background field was probided by a third channel. using a tungsten iodide source and appropriate filters. A broad band filter (Kodak 47B1 was used in the background to depress the short-waLelength sensitive mechanism.
Calibrariom The interference and ND. fixed filters were calibrated using a Carey-14 recording densitometer. In addition, the neutral filters and the wedge were calibrated in position for each of the interference filters. For this task a United Detector Technology PIN IO photodiode was used. Three shortwavelength filters 400. 410, and 420 were three-cavit) filters (Ditric Optics) with half-band passes of 7-9 nm and visible side bands blocked to 10m6. The remaining interference filters (Schott) had half-band-widths of 10-15 nm.
165
of the fovea1 cone
The relative brated
ipectrul
b! Optronics
response: of the PIN Laboratories.
used to check the relative
radiance
filters on each experimental day. An estimate of retmnl illuminance
diode
The calibrated through
was calidiode ua>
the interference
was made placing
an
llford SEI exposure photometer in the plane of the artificial pupil ior the standard tield with the 578 interference filter. The obr.tinsd measurement of retinal illuminance was YOO td. The retinal illuminance of rhs background \\as ascertained first bb establishing a HFP match for the 178 filter to the snndard illuminance and then b> direct homochromatic brig.htncss matching of the bariable field to the upper portion ot the background. The retinal illuminance of the background uas 500 rd. Sllhjrcrs The Jeuteranopes and protanopes were obtained as a result of advertising in the Unitersit) of Chicago student newspaper. All sere males. The) ranged in age from 17-31 with an average age of35. .AII observers were screened using A0 HRR plates. Ishihars pwudoisochromaric plates. the Farnsworth Munxll loo-hue test. and the Nagel anomaloscope. OnIF those observers who could match in color and brightness both rhe red (670 nm) and the green (536 nm) to the )ello\\ (5S9nm) on the Sage1 anomaloscope u-ere accepted as dichromuts. All observers had normal corrected acuit> and used rheir oustomarb refractive correction during the stud!.
At ths {tart of each session ths observer adjusted the chin rest until his eye was centered on the artificial pupil. The obserxsr then dark adapted for 5 min. Hetcrochromatic flicker photometry was performed to a 540 nm standard for protanopes and to a 580 nm standard for deutsranopes. The test walelengths were presented in rnndom order once in each session. The observer adjusted the radiance of the inriable test wavelength until there was a minimal or absent sensation of Hicker. The skperimenter then adjusted the radiance of the test wavelength in small steps around the observer’s setting in order 10 establish a match range. The Hicker rate (X-27 Hz) was adjusted to maintain the match range at 0-O-I log unit of radiance. After an initial practice session. the procedure v.as repeated on three separate sessions. The data reported are the average data for the three sessions.
Rewir.~
Figure 3 shout the log relative energ! as a function ofwa\elcngth for four deuteranopes (louer panel). and five protanopes (upper panel,. The solid lines show the relative sensititities of the transformations given in Table 1. The deuteranopic points scatter widei) across the predicted line. The protanopic points tend to lie above the prediction. Inset in each panel is the densit! difference betueen an HFP match at 54Onm and 464nm at 0 and 9 . i.e.: an estimate of macular pigment densit! for each observer. The protanopes shorted less \ariabilit\. in their macular pigment, ;I tinding apparent also in Pitt’s (1935) data. The estimated macular pigment density at 464 run ranged from 0.136 to 0% in the deuteranopes and from 0.263 to 0375 in the protanopes. The estimation may well be an underestimate since according to Polqak ( 1941) the
points tend to cluster above the predictions made by the transformations. the transformations anti part Bv definition. ium&sities of the normai iuminositj; iunction. These part tuminmities are, of course, determined by Judd’s (1951) revised luminosity curve. The Judd (1951) revised luminosity curve fits normal NFP data well above 430nm. The corrections made by Judd were based on the data of Wald (19453 (absolute threshold), Weaver (1949), Thompson (1949). and Ishak (192) (brightness ~at~h~n~ at short-wavel~ng~hs~ and Gibson and Tyndah (19131 ~extrapolatjon~. We have noticed small but consistent deviakms of the HFP data of Coblenz and Emerson (19 171, Sperling ( 19581, and Wagner and ~oynton (1971) from Judd’s (1951) revised fuminosity function at wavelengths below 430 nm. The Judd { 1% If revised luminosity function underestimates HFP luminosity (average of studies cited above) at wavetengths 4OO-430 nm. the difference being 0.31 log unit at 400 nm. 0.29 at 410 nm, 0.20 at 420 nm and @15 at 430 nm. Figure j shows the average of the dichromatic data plotted as log relative quanta1 sensitivity as a function
Fig. 3. Log reiative energy as a function of wavelength for five protanopes performing hererochromatic flicker photometry to a 540 nm standard [panel (a)] and 4 dcuteranopes [panel (b)] performing heterochromatic flicker photometry to 8 580 nm standard. The solid lines are the relative sensitivities of the transformations given in Table i. Inset in each panel is a ~omparis& of the match of 540 nm to 464 nm made with foveai iiaation and in the 9” vertical meridian. Symbols for panei [a): Protanopie observer RS O; MS V; DP A: K.a 22;MG 0. Symbols for panel(b): ~uterano~ic observer M3 0: KC Cl: NW V; fC 0.
macular pigment may extend to 17’. Further, the procedure implies unpravable (and improbable) assumptions of the identity of optical density, distribution, and sensitivity of the cones at the fovea and 9’ vertical mer-
idian. The same problems however are evident in all psychophysical techniques of estimating macular pigment. The measurement does pravide us with a method ofreducing observer variability due to interindividual variation in macular pigment density. Figure 4 shows the data corrected for each observer to a macular pigment of cl,,, 0.53. using the Wyszecki and Stiles f 1967) spectral correction. The solid lines are again the transformations shown in Fig 3. They provide a reasonable description of the data except b&w 430nm. For wavehzngths below 430 nm the data
167
Spectral sensitivity of the fovea1 cone
Wovenumber.
cm“
-2.21 25.300
20,000 Wavcnumbef.
L5.000
cm-”
Fig. 5. Log relative quantai sensitivity as a function of wavenumber for protanopes [panel (a)] and deuteranopes [panel (b)]. The dashed lines show the relative sensitivities ofthe transfor~tionsofTable I corrected to a retinal level. The solid lines are the predicted visual pigment absorption coefficient as given in Fig. 1. The inset in each panei shows the difference of the psychophysical estimates from the predicted visual pigment absorption coeffcient. The solid line in the insets is a factor for added lens density, the dashed line on the inset of panel (b) is a factor for added macular pigment. Symbols: average data of Pitt A; Hsia and Graham El; and the present study 0; all corrected to a retinal level. The crosses on the inset of panel (b) represent the difference between the transformation and its predicted visual pigment absorption coefficient.
of wavenumber. The data were corrected for an average lens and macular pigmentation ((i,,, of 053) as tabulated in Wyszecki and Stiies (1967). We have added two extra sets of dichromatic data: (a) Pitt’s ( 1935) protanopic luminosity data were added since in ‘The Pokorny and Smith study allows a comparison of the three tasks (absolute thresholds. HFP to 5SOnm and HFP to 3OOO’K)for five observers (two protanopes and three deutzranopes). The flicker rate below 500 nm was 10I2 Hz at 40 td. There was no difference in the thresholds as a function of method. The data are not replotted for reasons including: (a) we have no independent assessment of macular pigment and too few observers to justify averaging; (b) we could not measure sensitivities at wavelengths betow 420 nm: (c) examination of the long-wavelength slope suggests that our calibration procedure was not as precise as that used for this study.
our laboratory we can detect no difference in high and low frequency HFP functions for protanopes. ib) The Hsia and Graham (1957) absolute threshold data were added since Pokorny and Smith (1972) reported that using the Hsia and Graham (1957) procedure, absolute thresholds yielded identical results to HFP for dichromats.’ although other procedures (Guth and Lodge, 1973) do not show this identity. The Ikeda (1963, 1964) data suggest that stimulus duration is the determining parameter. The data for the Hsia and Graham (1952. 1957) and Graham and Hsia (1969) procedure (4 msec stimulus flash) suggest that their absolute thresholds follow the absorption coefficient of the most sensitive of the visual pigments. For protanopes and deuteranopes their respective long-wavelength-sensitive pigments appear to be the most sensitive over the full wavelength range, and may thus be compared to the same pigments isolated in the achromatic system by high-flicker rate HFP. The dashed lines show the relative sensitivity of the transformations of Table I corrected to a retinal level. The soIid lines are the proposed visual pigment absorption spectra based on the shape of the iodopsin absorption spectrum (Wald. Brown and Smith, 1955) calculated for appropriate optical densities. The spectra are slid on the horizontal axis to provide a good fit to the data and transformations on the long-wavelength slope with the result that the spectrum compared with the protanopic functions has a L,,, CCI. 532 nm and that compared with the deuteranopic functions has a i.,,, cu. S55 nm. These estimates for i,, are relative to the accuracy in i,,, of the iodopsin spectrum. The data points and the transformations fit the proposed visual pigment spectrum for wavelengths above 520 for deuteranopes and above 460 nm for the protanopes. There is a notch at 490 nm in the data, evident for both classes of observer, which may be an overcorrection of macular pigment at that wavelength (cf. vos, 1972). The deviations of the dichromatic data and functions from the proposed visual pigment spectrum become severe at short-wavelengths. inset in the graphs are the density differences of the dichromatic data points from the proposed visual pig.ment spectrum. The solid line through these points IS a correction of the lens density factor to 1.333 times the Wyszecki and Stiles (1967) tabulations. The solid line provides a good fit of the protanopic deviations. but not of the deuteranopic deviations. The dashed line on the deuteranopic inset which does provide a reasonable fit to the deuteranopic deviations is obtained by adding increased macular pigment of 0.12 optical density at 460 nm to the lens correction. DISCUSStOS
The results show that the dichromatic data and the transformation of the CIE data in the wavelength range 43~7~nm approximate the visual pigment absorption coefficient predicted from the iodopsin
Jle-waxiength cide a good iii TV> the Wri@t (1351) tritanopic coefficients. The c~~etti&nts at 520 nm were checked at :I number af values +I :tddrd macular pi~msnt at 460 nm. The preof,1 dictions fX within U.right’\ t 19521obxeriers proI iJed the 4en4t! \rf mitcuktr pigment Aied to the tongwa~elen&t ~2nsirii~ pighrtnr ha5 optical densit! at I60 nm betuecn 0419 timi 0.16. We thus conclude that the io&)psin spectrum when ndjuqted to appropriate densit) and slid MI the wa\enumber axis can pro\ ide a re;tsonable description of the L~YAJ:II pipmrnt absorption spectra of the mAidleand long-ua\elen$h sensitiie fove:tl cones although SC cspcct min~x deviations .tt the short-walelengths. Further. we come to the conclusion that we must ahon ditl’erent r\rntxints of macul;tr pigment for the two
Fig. 6. Panel (a) Log relative quanta1 sensitivity of the proposed human visual photopigments. The symbafs and lines correspond to the continuous lines of Fig, 5. Symbols: Middle-wavelength sensitive 0; long-wavelength sensitive a. Panel (b) Tritanopic coefficients predicted by the pigment absorption spectra of panel (a). The symbols are the same as those in Fig. I(b).
absorption spectrum (,at appropriate opticat densit!) prl” idsd the lens correttion factor is increased 01; RI. I.333 and that ;I diirerrtntial macular pigment correction is applied to the middle- and lon~wavelen~th sensitive mechanisms. There is no indic~ltion that tluorescence of the optic media or retina has contribrtted to
the measured sensitivities. The data do not show any heightened sensitivity that might be caused b! Huorescence. The added optic media correction factor is that predicted b\ C’oren and Girgus (1971) for JO-.vr-olds and agrees with Tan’s (1971) data for the ape group X-19. Although such a correction is necessarily CR/hoc, it is a reasonable correction. given our knowledge of the absorption in the ocular media. The macular pigment correction presents a different problem: it appears that the deuteranopes have slighti! sreater average macular pigment. The same result oc’curs in the transform~~tion of the revised CIE color mixture data, a result which preserves the inteprirr of the copunctal points. The long-wavetength sen4&e cones of normal trichromats are apparently subject to a 4ightly higher macular pigment than the mid-
Fig. 7. The effect of adding a selective Biter to the proposed Lang-~vave~~n~th sensitive pigment. Panel (a) Pigment absorption curves of Fig. 6 with a macular pigment filter of o.d. @il at 460 nm screening the long-wavelength sensitive photopigment. Panel (b) Ttitanopic coef%5enrs predicted by the spectral sensitivities shown in PaneI (a). The symbols are the same as those in Fig. i(b).
Spectral sensitivity of the iovcal cone
classes of cones: namely, maximal optical density of about 0.5 for middle-wavelength sensitive and @62 for long-wavelength icnsitice cones. This suggestion is not contradicted by the available psychoph>&tl measurements on normal macular pigment. since the predicted appxent average macular pigment density in normals given d,,,, of 0.5 for middle- and 0.61 for long-wavelength \sn\itivs cones would be about 057 at 460 nm. However. since the macular pigment is described as lying around the fibers of Henle (Segal, 1950).primarily in the outer plexiform layers (Gass, 1973). or through layers 4-9 of the retina (Polyak. 1941) the assumption of differential pigment density affecting the two classes of cones i4 somewhat puzzling. A number of alternatives may be considered. (a) A larger differential optical density of the two photopigments: In order to fit the deutan data to a hypothetical pigment absorption curve, we would need to compute it for dilute optical density. This possibility is contradicted by the results of Miller (1972) and ourselves (Smith and Pokorny. 1973) who find the longwavelength sensitive pigment to be in density 04-0.55. Further. our calculations of predicted tritanopic coefficients based on a larger optical density differential between the two pigments do not fit the Wright (1952) tritanopic data below 440 nm. (b) Two pigments in longwavelength sensitive cones: Two pigments each absorbing quanta should look like a pigment with apparently greater sensitivity than the proposed absorption spectrum. We are concerned with a sensitivity which lies below the proposed absorption spectrum. (c) Intrusion of the short-wavelength mechanism by inhibitory interaction with the long-wavelength mechanism. Such a hypothesis requires that some proportion of normal short-wavelength sensitive receptors exist whose output consists of an inhibitory linkage to the long wavelength sensitive receptors. These shortwavelength sensitive receptors would also be postulated for tritanopia-a possibility since there is no conclusive evidence that tritanopia is a pigment defect. However, our calculations show that predicted tritanopit coefficients for such a system do not agree with Wrights (1952) tritanopic data. (d) A stable photoproduct of the long-wavelength sensitive pigment acting as a screening pigment: This notion was suggested by Dartnall(1962) for the case of indicator yellow as a screening pigment for rhodopsin. The concept was carefully examined by Goldstein and Williams (1966) who concluded that the low optical densities in human rods and cones preclude stable photoproducts from acting as effective screening pigments. They worked however with Rushton’s (1956) value of 0.15 optical density for rods. We allow higher optical densities of 0.3-04. Even so, a substantial bleach (probably- 50 per cent) would be required to build up an appropriate density of any hypothetical screening photoproduct for long-wavelength sensitive cones. even given optimal conditions for such build-up. Our analysis indicates that the selective filtering of the
169
long-wavelength sensitive cones occurs at luminances ranging from absolute fovea1 threshold to SO0td. (e) .-\ substance of spectral shape similar to the macuhr pigment which is present in the structure of the cones and mav occur in both peripheral and fovea1 cones: The relatively higher dcnsitv of such a substance in the long-wavelength sensitive cones may reflect a greater length of the outer segment, a result which we might expect based on the higher density of the long-wavelength sensitive cones found by Miller (1972) and Smith and Pokornl (1973). In their estimation of the optical density ol’ human cones. Dobelle. Marks and YlacXchol (1969) imply that the optical densit) ivill be proportional to the length of the outer segment. (f) There is the possibility of error in the Wald. Brown and Smith ( 1955) measurements. At 500 nm the percentage extinction for iodopsin is 50 per cent and we would need to invoke a IO per cent error in Wald. Brown and Smiths (1955) spectrum. The two estimates using independent techniques made by Wald. Brown and Smith ( 1955) of iodopsin absorption at 500 nm differed by only 2 per cent. (g) A remaining alternative would be to state that the deuteranopic curve is the true absorption curve and that the protanopic curve results from a summation such as mentioned in (b) above. Such a hypothesis predicts incorrect tritanopic coefficients. Further we would need to accept that the similarity in shape of the protanopic functions to the iodopsin spectrum is coincidence. We prefer to think that the shape similarities found between the visual pigments of many species in the animal kingdom does extend to the human cone photopigments. Of these alternatives, the most likely seems to be (d) and (e). which of course are essentially ad ltoc explanations. Thus vve come to the conclusion that our analysis requires a selective screening of long-wavelength sensitive comes by a filter whose spectral transmisston is similar or identical to macular pigment. Guild (1931). Wright (1923) and Stiles (1955) found considerable variab&ty- in the color matches of normal trichromats for the spectral region 470-500 nm. W’right (19X3) points out the different color matches esceed observdtional error and may indeed reflect receptor differences. From our point of view such variability might reflect variation in differential density of the screening pigment. Further. it appears that the human cone visual pigments are closely similar in shape to the Wald Brown and Smith (1955) measured absorption spectrum for iodopsin. However we would not be surprised by minor deviations in the spectral region 400450nm. the region of greatest uncertainty in both Wald Brown and Smiths (1955) data and the psychophysics. We conclude that the shape of the iodopsin absorption spectrum provides a basis for calculation of absorption spectra for the middle- and long-wavelength sensitive human cone photopigments.
REFEREhCES
Xlpcrn 41.. Xlaaseidvaag F. and Ohba N. ( IYTI) The kine tics of cone visual pigments in man. t’isioft Rrs. 11. j39549. Boettner E. A. and VVoltet 5. R. I 1962)Transmission of the ocular media. Inrrsr. Ophthai. 1, 776-783. Bone R. A. and Spsrrock J. 51. B. (1971J Comparison of macular pigment densities in human eyes. C’isio/tR~s. 11. IOj7- 106-t. Brindle! G. S.. Du Croz J. .I. 3rd Rushton W’.A. H. (19661 The tlicker fusion frequent)- of the blue-sensitive mechanism of colour vision. j. Ph_Mffi, Lond. 183, 497500. Brolin S. E. and Cederlund C. (1953) The fluorescence of the lens of the eves of different species. Actrr ophrhal. 36, 37c32s _
.
Coblentz W. W. and Emerson W. B. (1915) Relative sensibilitv ofthc average eve to lizht of different coIors and some p&tical applications to radiation problems. Bttil. n~trz. Bur. Sron&rinms1-l. 167-236. Collins F. D., Love R. M. and \lorton R. A. i 1952) Studies in rhodopsin-IV: Preparation of rhodopsin. Biochrm. J. ;I, 192-29s. Coren S. and Girgus J. S. (1972). Density of human lens pigmentation: irr rirr) measures over an extended age ranye. t-isiuct Rcs. 12, %.r-316. Dartnail H. J. A. 113621 The photobiology of Visuai Processes. In Tile Ertrdited by Davson H.), Vol. II. pp. ??I533. Academic Press, New York. Dobellc W. H.. Xlxks W. B. and MacNichol E. F. (19691 Visual pigment density in single primate fovea1 cones. Scit%ce. .\‘. y. 166. IbOS-I 510. Gass 1. D. M. (19-2) Nicotinic acid macttlopathy. .-lrrr.J. ~~~ir~~~~. 76. SO&5 Ii). Gibson K. S. and Tyndall E. P. T. i 1923) Visibility of radiant energy. Scienr. PHI?.Birr. Srund. 475, 131-191. Goldstein E. B. and WXiams T. P. (1966) Calculated effects of “screening pipm:nts”. Vision Res. 6, 39-50. Graham C. I-f. and Hsia Y. (1969) Saturation and the fovcal achromatic interval. d. opr. SM. .Arn.59,993-997. Green D. G. (1969) Sinusoidal fiicker characteristics of the color-sensitive mechanisms of the eye, Vision Rrs. 9, 591601. Griitzner P. and Kohlrausch A. (1961) Der lichtwrlust in der macula lutsa und das farbensehen von deuteranomalen. PfIiigerr Arch. ges. Ph,vsiof. 274, 3 I S-330. Guild J. (1931) The calorimetric properties of the spectrum. Phi!. Trtrrts. R. SOC. 4 230, 149 157. Guth S. L.. Aieuander J. V.. Chumbly J. I.. Gillman C. B. and Patterson hf. 41. (196s) Factors affecting luminance additivity at threshold among normal and color-blind subjects and elaborations of a trichromatic-opponent colors theory. I-isinn Res. 8, 913-925. Guth S. L. and Lodge H. R. (19733Heterochromatic additivity. fovenl spectral sensitivity and a new color model. f. opt. Sac. .-bn. 63. -tit%462. Hsia Y. .rnd Graham C. Ii. (1953)Spectral sensitivity of the cones in the dark adapted human eye. Proc. mm. Acud. Sci. 38, SC-Sj. Hsia Y. and Graham C. H. (1957) Spectral luminosity cur\es for nrotanooic. deuteranopic and normal subjects. Proc. n40;. ilcad. Sci. 13. 101I-1019. Ikeda M. (1963) Study of interrelations between mechanisms at threshold. J. opt. Sue. Am. 53, 130j_f313. fkcda M. (1964) Further use of the summation index for the study of color vision. J. opt. Sot. ,Arrz.S4, 89-9-k
Ishak I. G. H. I ii)jZ) The photopic lummosity curve for a group oftiftsrn Egyptian trtchromats. J. clpr. SOC. im. 42. 519-j;!,
Judd D. 3. I 195:JSeer. Rep. Coiorimetry and artificial daylight. In Proc. 13~11Srssip~t CIE: Stockholm Vol. I. Tschn. Comm.. No. -. Klanp G. (1%) Measurements and studies of the Huorsscrnce of the human lens in ciro. .&m ophthctl. Suppl. 31. I-Ii. Lr Grand Y. I IY3S) Sur la tiuorsscence du cristailin. C.r+. .k‘Zif. Sci. zu:.
11zs- I 130.
L. Grand Y. (19fi) Nou~clles rzcherchrs sur la fluorescence du cristallin. C.r. ,fcnd. Sci. 225. 1031-10~2. Ludvigh E. and McCarthy E. F. (1939) Absorption of visible light by refractive media of the human eye. .-l~hs. Opitthai. 20. zi-51.
Miller S. S. (1973)Psychophysical estimates of visual pig. menr densities in red-gcen dichromats. J. Plt.~siof.. Lmf. 223. Y9- 107. Pitt F. H. G. (1935) Characteristics of dichromatic vision, with an appendix on anomalous trichromatic vision. Spec. Rep. wd.
Rex Council 200, I-3.
Pokorny J. and Smith V. C. (1972) Luminosity and CFF in oenteranopss and protanopes. J. opt. Sot. .irn. 62, 1I I-1 17. Polyak S. L. (IY-IL)The Rrritra. L’niversitv of Chicago. Ruddock K. H. 11963) Evidence for macttlx pi~menmtion from colour matchine data. lisi~n Rzs. 3. ili-429. Rushton Et’. ,A. H. (1956) The Rhodopsin density in the human rods. J. Ph,vsiof.. Lad. 134, 306. Said F. S. and Weals R. A. (1959) The variation with age 01 the spectral transmissivity of the living human crystalline lens. Grr-ortro/ogici3, 3 I i-33 I. Ssgal J. I 1950)Localisation du pigment maculairs de la rCrinr. Cx. &rrc. Sot. &of. I-M, Pt. 2, 163&1631. Smi:h V. C. and Pokorny J 11972) Spectral sensitivity of color-blind observers and the cone photopigments. t’isioh Rrs. 1’-9 203-2071.
Smith V. C. and Pokornl J. (1973) Psychophysical estimates of optical densit\ in human cones, C’isiortRrs. 13, I l99-
12oi. Sperling H. G. I 19%) An experimental investigation of the relationship between colour mixture and luminous et%c&x,-. In Prot. Yrh !VPL Prohte,ns 0s C&t-. (Symposium held at the Sarional Physical Laboratory. September 1957) pp. 251-‘79. Her Majesty’s Stationery Office, London. Stiles IV. S. t 1955) Interim report to the commicsion intsrnationale de l’eclarrage, Zurich. 1955, on the Xational Physical Laboratory’s investigation of ~olour-matching (1954). with an appendix by W. S. Stiles and J. 41. Burch. Dprica .Jcm 2, 16%ISI. Stiles VV.S. and Burch J. >I. (1939) N.P.L. colour-matchinp in\sstigation: final report (19SS). Opricu -Icr,t 6, l-26. Tan K. E. W. P. (1971) Vision in the ultraviolet. Doctoral thests. Utrecht. Thomson L. C. 119-19)Shape irregularities in the equal energy- Ium~nosity curve. Prof. Phys. Sot.. fmd. B62. 757?I?,
Thomson L. C. and Wright IV. 5 (1953) The convergence of the tritanopic confusion loci and the derivation of the fundamental response functions. J. opt. Sot. .-lm.13, 890s94. Trendelenberg W. E. T. (1961) Der Grsichrsinrt. Grund:ii~gr der ph~siologi~c~re}t opiik. 2nd Edition (edited by Monje M.. Schmidt I. and Schlitz E.). pp. 67-69. Sprinpr-Veriag, Berlin.
Spectral sensitivity of the fovea1 cone Vos J. J. (1973) Literature review of human macular absorption in the visible and its consequences for the cone receptor primaries. Inst. Percept. RVO-TN0 Report So. 1ZF (972-l 7. Soesterberg. Vos J. J. and Walraven P. L. (1971) On the derivation of the fovea1 receptor primaries. Vision Res. 11. 799-818. Wagner G. and Boynton R. M. (1972) Comparison of four methods of heterochromatic photometry. J. opt. Sot. .-lm. 62, IjOS-1515. Wald G. (1945) Human vision and the spectrum. Science, S.l’: 101. 653-658. Wald G.. Brown P. K. and Smith P. H. (1955) Iodopsin. J. gen. PhysioL 38, 623-681. Weale R. A. (1954) Light absorption by the lens of the human eye. Optical .-lcta 1. 107-l 10.
171
Weaver K. S. (L949)A provisional standard observer for low level photometry. J. opt. Sot. ,4/n. 39, 278-39 1. Willmer E. N. and Wright W. D. (1945) Colour sensitivity of the fovea centralis. .Vaturr. Land. 156, 119-L21. Wright W. D. (1929) A re-determination of the trichromatic coefficients of the spectral colours. Trans. opt. Sot. U.K. 30, 141-164. Wright W. D. (1951) The visual sensitivity of normal and aphakic observers in the ultra-violet. Ann. Ps@ol. 50, 169-177. Wright W. D. (1952) The characteristics of tritanopia. J. opt. Sot. Am. 12. 5G9-521. Wyszecki G. and Stiles W. S. (1967) Color Science: Concepts and Methods, Quantitative Data and F~rmwlas. John Wiley, New York.
